What problem does it solve? Autonomous workflow runs in Nexi's agency engine can get stuck in pending, stop reasoning, misreport progress to the UI, or lose their history, and this Skill provides the mental model and debugging checklist to pinpoint whether the cause is autonomy gating, event wiring, persistence, or an actual bug. ## Core Features & Use Cases - Pass Pipeline Map: Documents the Planner, Research, ToolOperator, Verifier, Reflection, and Report passes implemented as (run, ctx) functions in workflow_engine.py, so you upgrade a pass in place instead of restructuring. - Autonomy Gating Reference: Explains NEXI_AGENCY_AUTONOMY and NEXI_AGENCY_MODE (locked | manual | supervised | autonomous_safe) so you check intended gates before assuming a crash. - Run Lifecycle & Persistence: Covers WorkflowRun status transitions, JSONL persistence at data/nexi/agency/workflow_runs.jsonl, register_workflow_type(), and the EVENT_SINK HUD hook. - Use Case: A run appears stuck in pending — instead of digging through the runner, you check autonomy_mode() first and discover the mode is manual, meaning the run is intentionally waiting for a human decision. ## Quick Start Ask the assistant to debug why a Nexi agency workflow run is stuck in pending and not reasoning, using the workflow engine's autonomy flags and event sink wiring.
